torchembed 0.3.0

A PyTorch library of modern embedding strategies missing from torch.nn

torchembed

Modern embedding strategies for PyTorch — the ones missing from torch.nn.

torch.nn gives you nn.Embedding (a lookup table). That's it. The moment you work with continuous inputs, modern transformer architectures, coordinates, time, or tabular data, you're on your own — copy-pasting RoPE implementations across projects.

torchembed is a single, well-tested, pip-installable home for all of them.

---

Table of Contents

• Installation
• What's included
• Quick start
• Triton kernels
• Documentation
• Examples
  • Rotary Embedding (RoPE)
  • ALiBi
  • Gaussian Fourier Projection
  • ViT Patch Embedding
  • Tubelet Embedding
  • Tabular categorical features
  • Cyclic time features
  • Frequency Embedding
  • Random Fourier Features
• Design principles
• Running tests
• Contributing
• License

---

Installation

pip install torchembed

For GPU-accelerated kernels:

pip install torchembed[triton]

Requires Python >= 3.9 and PyTorch >= 2.0.

---

What's included

Module	Class	Use case
positional	RotaryEmbedding	Modern LLMs (LLaMA, Mistral, Falcon)
positional	ALiBiEmbedding	Long-context models (BLOOM, MPT)
positional	SinusoidalEmbedding	Classic Transformers
positional	LearnedPositionalEmbedding	BERT, GPT-2
fourier	RandomFourierFeatures	Kernel approximation, coordinate encoding
fourier	LearnedFourierFeatures	Trainable frequency decomposition
fourier	GaussianFourierProjection	Diffusion models (timestep embedding)
categorical	EntityEmbedding	Tabular categorical features
categorical	MultiCategoricalEmbedding	Multiple categorical columns at once
patch	PatchEmbedding	Vision Transformers (ViT)
patch	TubeletEmbedding	Video Transformers (VideoMAE, ViViT)
temporal	CyclicEmbedding	Hour, day, month (cyclic features)
temporal	TimestampEmbedding	Continuous timestamps
temporal	FrequencyEmbedding	Time series, periodic signals

---

Quick start

from torchembed.positional import RotaryEmbedding
from torchembed.fourier import GaussianFourierProjection
from torchembed.patch import PatchEmbedding

rope = RotaryEmbedding(dim=64)
q_rot, k_rot = rope(q, k)

t_emb = GaussianFourierProjection(embed_dim=256)
emb = t_emb(t)

patch_emb = PatchEmbedding(image_size=224, patch_size=16, embed_dim=768)
tokens = patch_emb(images)

---

Triton kernels

torchembed includes optional triton-accelerated kernels for GPU. Install with:

pip install torchembed[triton]

Enable with use_fused=True:

rope = RotaryEmbedding(dim=64, use_fused=True)

The fused RoPE kernel combines cos/sin lookup, rotate-half, and element-wise multiplication into a single triton launch, reducing memory traffic. Supports any even dim (32, 64, 128, etc.) and full autograd support. Falls back to vanilla PyTorch automatically when triton is unavailable or inputs are on CPU.

Benchmarks

RoPE forward pass on NVIDIA GB10 (float16):

Shape (B,H,S,D,rot)	PyTorch (ms)	torch.compile (ms)	Triton (ms)	Speedup
(1,32,2048,128,128)	1.40	0.61	0.34	4.15x
(1,32,4096,128,128)	2.95	1.21	0.63	4.68x
(1,32,8192,128,128)	5.94	2.47	1.29	4.62x
(2,32,2048,128,128)	2.97	1.23	0.75	3.98x
(1,32,2048,256,128)	2.87	1.24	0.66	4.34x

The fused Triton kernel is ~4x faster than pure PyTorch and ~2x faster than torch.compile. torch.compile reduces overhead but cannot eliminate intermediate tensor allocations from chunk/cat — the fused kernel reads and writes each element exactly once.

---

Documentation

Full API reference for every module is in the docs/ directory:

Module	File
Positional (RoPE, ALiBi, Sinusoidal, Learned)	docs/positional.md
Fourier features	docs/fourier.md
Categorical embeddings	docs/categorical.md
Patch embeddings (ViT, video)	docs/patch.md
Temporal embeddings	docs/temporal.md

---

Examples

Rotary Embedding (RoPE) — LLaMA / Mistral style

import torch
from torchembed.positional import RotaryEmbedding

rope = RotaryEmbedding(dim=64)  # head_dim

# Inside your attention layer:
q = torch.randn(batch, heads, seq_len, 64)
k = torch.randn(batch, heads, seq_len, 64)
q, k = rope(q, k)  # apply rotation in-place

RoPE has no trainable parameters and preserves vector norms (it's a pure rotation). The default base of 10,000 matches the original paper; use base=500_000 for LLaMA 3.

For GPU-accelerated inference:

rope = RotaryEmbedding(dim=128, use_fused=True).to("cuda")
q, k = rope(q.cuda(), k.cuda())

---

ALiBi — long context with length extrapolation

from torchembed.positional import ALiBiEmbedding

alibi = ALiBiEmbedding(num_heads=8)

# After computing raw attention scores:
attn_scores = q @ k.transpose(-2, -1) / math.sqrt(head_dim)
attn_scores = alibi(attn_scores)   # adds learned distance penalty
attn_weights = attn_scores.softmax(-1)

---

Gaussian Fourier Projection — diffusion model timestep embedding

from torchembed.fourier import GaussianFourierProjection
import torch.nn as nn

class DiffusionTimeEmbedding(nn.Module):
    def __init__(self, embed_dim):
        super().__init__()
        self.fourier = GaussianFourierProjection(embed_dim=embed_dim, scale=16)
        self.mlp = nn.Sequential(
            nn.Linear(embed_dim, embed_dim * 4),
            nn.SiLU(),
            nn.Linear(embed_dim * 4, embed_dim),
        )

    def forward(self, t):
        return self.mlp(self.fourier(t))

t_emb = DiffusionTimeEmbedding(embed_dim=256)
t = torch.rand(32)   # normalized timesteps
emb = t_emb(t)       # (32, 256) — condition your UNet on this

---

ViT Patch Embedding

from torchembed.patch import PatchEmbedding

patch_emb = PatchEmbedding(
    image_size=224,
    patch_size=16,
    embed_dim=768,
)

images = torch.randn(4, 3, 224, 224)
tokens = patch_emb(images)    # (4, 196, 768)
print(patch_emb.num_patches)  # 196

---

Tubelet Embedding — Video Transformers

from torchembed.patch import TubeletEmbedding

tubelet_emb = TubeletEmbedding(
    image_size=224,
    patch_size=16,
    tubelet_size=2,
    embed_dim=768,
)

video = torch.randn(2, 3, 16, 224, 224)   # (B, C, T, H, W)
tokens = tubelet_emb(video)                # (2, 1568, 768)
# 1568 = (16/2) * (224/16) * (224/16) = 8 * 14 * 14

---

Tabular categorical features

from torchembed.categorical import MultiCategoricalEmbedding

# A tabular dataset with 3 categorical columns:
# country (50 unique values), day of week (7), product category (120)
emb = MultiCategoricalEmbedding(cardinalities=[50, 7, 120])
print(emb.output_dim)   # sum of auto-sized embed dims

x = torch.stack([country_ids, dow_ids, category_ids], dim=1)   # (batch, 3)
features = emb(x)   # (batch, output_dim)

---

Cyclic time features

from torchembed.temporal import CyclicEmbedding
import torch

hour_enc  = CyclicEmbedding(period=24)
dow_enc   = CyclicEmbedding(period=7)
month_enc = CyclicEmbedding(period=12)

hour   = torch.tensor([0.0, 6.0, 12.0, 18.0])
dow    = torch.tensor([0.0, 1.0, 2.0, 3.0])
month  = torch.tensor([1.0, 4.0, 7.0, 10.0])

time_features = torch.cat([
    hour_enc(hour),    # (4, 2)
    dow_enc(dow),      # (4, 2)
    month_enc(month),  # (4, 2)
], dim=-1)             # (4, 6)

---

Random Fourier Features for coordinate encoding

from torchembed.fourier import RandomFourierFeatures

# Encode 2D spatial coordinates for a neural field / NeRF-style model
rff = RandomFourierFeatures(in_features=2, out_features=256, sigma=1.0)

coords = torch.rand(1024, 2)   # (x, y) pairs in [0, 1]
features = rff(coords)          # (1024, 256)

---

Frequency Embedding — learnable periodic decomposition

from torchembed.temporal import FrequencyEmbedding

# Discover periodic structure in time series automatically
freq_emb = FrequencyEmbedding(embed_dim=32)

t = torch.linspace(0, 100, 512).unsqueeze(0)   # (1, 512) time steps
out = freq_emb(t)                               # (1, 512, 33)
# 33 = 1 linear trend + 32 sinusoidal components

---

Design principles

Everything is an nn.Module. You can use any embedding as a layer in a larger model, save/load it with state_dict, move it across devices, and wrap it with torch.compile.

No required dependencies beyond PyTorch. torchembed has exactly one required dependency: PyTorch itself. We don't pull in transformers, numpy, or anything else. Triton-based GPU kernels are optional (pip install torchembed[triton]).

Device-agnostic. No .cuda() calls inside the library. Move your model to whatever device you want — the embeddings follow.

Bring just what you need. Every embedding class is independent. Use one, use all, use none — no framework lock-in.

---

Running tests

pip install torchembed[dev]
pytest

Building API docs

pip install torchembed[dev]
make docs    # generates docs/api/
make docs-serve  # serves at http://localhost:8080

API docs are generated from docstrings using pdoc. The hand-written guides in docs/ complement the API reference. Source code uses Google-style docstrings.

---

Contributing

Contributions welcome! If there's an embedding strategy you find yourself copy-pasting into projects, open a PR. Please include:

• The module with a clear docstring and paper reference
• Tests covering shape, gradients, and key mathematical properties
• An example in the README

---

License

MIT